search

MathHubInfo / Legacy-localmh-docker

Legacy docker wrapper for lmh
1 stars 1 forks source link

Missing package and (partial) cause of slowness ? #18

Open m-iancu opened 9 years ago

m-iancu commented 9 years ago

On the latest version, with updated latexml, I get a lot of errors "HTTP GET on https://mathhub.info/smglom/algebra/pgfutil-latex.def.tex failed. Reason: Can't connect to mathhub.info:443 (Connection timed out) "

We had an e-mail discussion about that a while back (about problems with german articles and babel) with @dginev and @kohlhase and it seems that the package liblwp-protocol-https-perl needs to be installed. It's not completely clear why it is actually needed or why it helps but it does. I manually installed it locally (inside docker) and I don't get the errors anymore.

It also seems faster. The last run over smglom took (for me) a whopping 42 hours for ~1800 files which means less than 45 per hour. After installing the package it did 120 in less than 10 minutes which is more reasonable. I didn't finish running yet though, so I'm not sure if this really fixes it (maybe its just that some files are just much slower than others).

tkw1536 commented 9 years ago

Committed the fix, I am rebuilding the docker image right now. I will push that tonight.

dginev commented 9 years ago

It's not completely clear why it is actually needed or why it helps but it does.

To make it clear: While working on sTeX-from-the-web workflows with the LaTeXML fork, we needed an https-capable fetch for LaTeXML, which is why the package is needed - it gives https capabilities to Perl's standard LWP module.

Especially given that the error you're showing is exactly failed GETs on https, it's pretty clear how the installed package will patch it.

m-iancu commented 9 years ago

@dginev indeed:

  1. I should have said completely clear to me
  2. Last time we discussed this, the package was found locally by LaTeXML so it wasn't clear why the https-fetching default case was even reached. As far as I understand it shouldn't be reached here either (assuming lmh does install and setup LaTeXML & friends properly). But of course, I understand LaTeXML/sTeX very little :)

Either way, it works now (on my local intallation), a full run finished in 2 hrs now so 20 times faster, thats pretty good. I haven't tried Tom's new image yet but that should work too.